Predicting person´s Zheng states using the heterogeneous sensor data by the semi-subjective teaching of TCM doctors

This paper proposes a method for predicting person´s Zheng states of traditional Chinese medicine (TCM) using the heterogeneous patterns of samples which are acquired from the multi-sensors and semi-subjectively labelled by TCM doctors (TCMDs). After confirming the difference from the TMDs´ diagnosis and the overlapping of samples belonging to the Zheng class and its negative class based on the extracted eigen attributes, an index called separation for measuring the separability of these two classes is defined to investigate the relationship with the Zheng scores. On the basis of that, the scheme of generating the Zheng´s classifier and predicting the person´s Zheng states is described. Further, the relation of the separation with the performance of the prediction is analyzed. The experimental results illustrate that the proposed approach is feasible.